Dynamic Security - Assessment

4.
Lyapunov Stability
Nonlinear Control Systems
António Pedro Aguiar

pedro@isr.ist.utl.pt
4. Lyapunov Stability
IST DEEC PhD Course

http://users.isr.ist.utl.pt/%7Epedro/NCS2012/
2012
1
Autonomous System
Consider the autonomous system
ẋ = f (x) (1)
where f : D → Rn is a locally Lipschitz map from a domain D ⊂ Rn into Rn and
there is at least one equilibrium point x̄, that is f (x̄) = 0.
Goal: Stability analysis of the equilibrium point x̄ ∈ D.
Without loss of generality, we consider that x̄ = 0.

Why? If it is not, then consider the change of variables y = x − x̄. Then
ẏ = f (x) = f (y + x̄) := g(y)
where g(0) = 0
Definition
The equilibrium point x = 0 of (1) is
• stable if, for each > 0, there is δ = δ() > 0 such that
kx(0)k < δ ⇒ kx(t)k < , ∀t ≥ 0
• unstable if it is not stable

• asymptotically stable if it is stable and δ can be chosen such that
kx(0)k < δ ⇒ lim x(t) = 0

t→∞
2
Example - Pendulum
ẋ1 = x2
ẋ2 = − gl sin (x1 ) − β
x
m 2
where x1 = θ, x2 = θ̇, and β is the friction. The following equilibrium points

• x = (0, 0) (which is a stable focus) is asymptotically stable
• x = (π, 0) (which is a saddle point) is unstable
For β = 0 then the above equilibrium points are stable (but not asymptotically).
3
A function V : D → R is said to be
• positive definite if V (0) = 0 and V (x) > 0, ∀x 6= 0
• positive semidefinite if V (0) = 0 and V (x) ≥ 0, ∀x 6= 0
• negative definite (resp. negative semi definite) if −V (x) is definite positive (resp.
definite semi positive).
In particular, for V (x) = xT P x (quadratic form), where P is a real symmetric matrix,
V (x) is positive (semi)definite if and only if all the eigenvalues of P are positive
(nonnegative), which is true if all leading principal minors of P are positive (all
principal minors of P are nonnegative).
Example
V (x) = ax21 + 2x1 x3 + ax22 + 4x2 x3 + ax23

2 32 3
a 0 1 x1
= x1 x2 x3 4 0 a 2 5 4 x2 5 = xT P x
ˆ ˜
1 2 a x3
˛ ˛
˛ a 0 ˛
The leading principal minors of P are a, ˛˛ ˛ = a2 , and
0 a ˛
˛ ˛
˛ a 0 1 ˛
˛ 0 a 2 ˛ = a3 − a − 4a = a a2 − 5
˛ ˛ ` ´
˛ ˛
˛ 1 2 a ˛
√
Therefore, V (x) > 0 if a > 5.
4
Lyapunov’s stability theorem
Theorem 4.1 - Lyapunov’s stability theorem

Let x = 0 be an equilibrium point for ẋ = f (x) and D ⊂ Rn be a domain containing
x = 0. Let V : D → R be a continuously differentiable function such that
• V (0) = 0, V (x) > 0, ∀x ∈ D\ {0}
• V̇ (x) ≤ 0, ∀x ∈ D
Then, x = 0 is stable. Moreover, if
V̇ (x) < 0, ∀x ∈ D\ {0}
then x = 0 is asymptotically stable.
5
Proof
• Given > 0, chose r ∈ (0, ] such that
Br = {x ∈ Rn : kxk ≤ r} ⊂ D
• Let α = minkxk=r V (x). Then α > 0.

Take β ∈ (0, α) and let
Ωβ = {x ∈ Br : V (x) ≤ β}
Then Ωβ ⊂ Br .
Why? Because if it was not, then there is a point p ∈ Ωβ that lies on the
boundary of Br . But V (p) ≥ α > β, although for all x ∈ Ωβ , V (x) ≤ β which is
a contradiction.
• The set Ωβ is an invariant set, that is, for any trajectory starting in Ωβ at t = 0
stays in Ωβ , ∀t ≥ 0.
Why? Because
V̇ (x(t)) ≤ 0 ⇒ V (x(t)) ≤ V (x(0)) ≤ β, ∀t ≥ 0
Note also that there exists a unique solution defined for all t ≥ 0 whenever
x(0) ∈ Ωβ because Ωβ is a compact set (closed and bounded since it is contained
in Br ).
6
Proof
• By continuity of V (x) and V (0) = 0, we conclude that there is a δ > 0 such that
kx(t)k ≤ δ ⇒ V (x) < β
Then,
Bδ ⊂ Ωβ ⊂ Br
and
x(0) ∈ Bδ ⇒ x(0) ∈ Ωβ ⇒ x(t) ∈ Ωβ ⇒ x(t) ∈ Br
Therefore,
kx(0)k < δ ⇒ kx(t)k < r ≤ , ∀t ≥ 0
which shows that the equilibrium point x = 0 is stable.
7
Proof
To prove asymptotically stability and assuming that V̇ (x) < 0 ∀x ∈ D\{0} we have to
show that x(t) → 0 as t → ∞, that is
∀a > 0 ∃τ > 0 : kx(t)k < a, ∀t > τ
But
∀a > 0 ∃b > 0 : Ωb ⊂ Ba
So it is sufficient to show that
V (x(t)) → 0 as t → ∞
Since V is monotonically decreasing and bounded from below by zero, then
V (x(t)) → c ≥ 0 as t → ∞
We need to show that c = 0. By contradiction, suppose that c > 0, which implies that
the trajectory lies outside the ball Bd ⊂ Ωc . But
Z t
V (x) = V (x(0)) + V̇ (x(τ ))dτ ≤ V (x(0)) − γt
0
because V̇ (x(τ )) ≤ −γ = maxd≤kxk≤r V̇ (x). Thus, V will eventually become

negative, which is a contradiction since V > 0.
8
Examples
1.
ẋ = −g(x), x∈R
with g locally Lipschitz on (−a, a) and
g(0) = 0, xg(x) > 0, ∀x 6= 0, x ∈ (−a, a)
Let Z x
V (x) = g(y)dy
0
It is continuously differentiable and positive definite. Thus, V (x) is a valid
Lyapunov function candidate. To check if it is a Lyapunov function, we compute
∂V
V̇ (x) = − g(x) = −g(x)2 < 0, ∀x ∈ D\{0}
∂x
Hence, the origin is asymptotically stable.
9
Examples
2. Consider the Pendulum example without friction
ẋ1 = x2
ẋ2 = −a sin (x1 )
Assume the following energy function

Z x1
1 1
V (x) = a sin(y)dy + x22 = a(1 − cos(x1 )) + x22 .
0 2 2
Clearly, V (0) = 0 and V (x) > 0, −2π < x1 < 2π, x1 6= 0.
V̇ (x) = a sin(x1 )ẋ1 + x2 ẋ2

= a sin(x1 )x2 − x2 a sin(x1 ) = 0
Thus the origin is stable. Since V̇ (x) = 0, we can also conclude that the origin is
not asymptotically stable.
10
Examples
3. Pendulum equation, but this time with friction
ẋ1 = x2
ẋ2 = −a sin (x1 ) − bx2
Consider
1 2
V (x) = a(1 − cos(x1 )) + x
2 2
Then,
V̇ (x) = −bx22 ≤ 0
which is negative semidefinite.
why? because V̇ (x) = 0 for x2 = 0 irrespective of the value of x1 .
We can only conclude that the origin is stable!
11
Examples
3. However, we know that is asymptotically stable. Let us try

1 T
V (x) = x P x + a(1 − cos(x1 ))
2
where P given by » –
P11 P12
P =
P12 P22
2 > 0.
is positive definite if P11 > 0 and P11 P22 − P12
Computing V̇ and taking P22 = 1, P11 = bP12 , P12 = b/2, yields
1 1
V̇ = − a b x1 sin(x1 ) − b x22
2 2
The term x1 sin(x1 ) > 0 for all 0 < |x1 | < π.
Taking D = {x ∈ R2 : |x1 | < π} we conclude that V is a Lyapunov function and
the origin is asymptotically stable.
This example emphasizes an important feature:
The Lyapunov theorem’s conditions are only sufficient!
12
Region of attraction
Let φ(t, x) be the solution of ẋ = f (x) that starts at initial state x at time t = 0.
Then, the region of attraction is defined as the set of all points x such that φ(t, x) is
defined for all t ≥ 0 and
lim φ(t, x) = 0.
t→∞
If the Lyapunov function satisfies the conditions of asymptotic stability over a domain
D, then the set
Ωc = {x ∈ Rn : V (x) ≤ c} ⊂ D
is an estimate of the region of attraction.
When the region of attraction is Rn ?

That is, when x = 0 is globally asymptotically stabe (GAS)?
Clearly D = Rn , but is this enought?
13
NO!
x21
V (x) = + x22
1 + x21
Ωc is unbounded with c large! For Ωc to be in the interior of a ball Br , c must satisfies
c < inf V (x).

kxk≥r
If
l = lim inf V (x) < ∞
r→∞ kxk≥r
then Ωc is bounded only if c < l. In the example
x21 x21
l = lim min [ 2
+ x22 ] = lim [ ]=1
r→∞ kxk=r 1 + x1 kx1 k→∞ 1 + x21
Thus Ωc is bounded only for c < 1. To ensure that Ωc is bounded for all values of
c > 0 we need the radially unbounded condition
V (x) → ∞ as kxk → ∞
14
Laypunov Stability - Globally Asymptotically Stability
Theorem 4.2 - GAS

Let x = 0 be an equilibrium point for ẋ = f (x). Let V : Rn → R be a continuously
differentiable function such that
• V (0) = 0 and V (x) > 0, ∀x 6= 0
• kxk → ∞ ⇒ V (x) → ∞
• V̇ (x) < 0, ∀x 6= 0.
Then x = 0 is globally asymptotically stable (GAS).
15
Instability Theorem
Theorem 4.3 - Instability Theorem

Let x = 0 be an equilibrium point for ẋ = f (x). Let V : D → R be a continuously
differentiable function such that V (0) = 0 and V (x0 ) > 0 for some x0 with arbitrarily
small kx0 k. Choose r > 0 such that the ball Br = {x ∈ Rn : kxk ≤ r} is contained in
D. Define a set U given by
U = {x ∈ Br : V (x) > 0}
and suppose that V̇ (x) > 0 in U . Then, x = 0 is unstable.
16
The Invariant Principle

Definitions
• P is said to be a positive limit point of x(t) if there is a sequence {tn }, with

tn → ∞ as x → ∞ such that x(tn ) → p as n → ∞.
• The set of all positive limit points is called positive limit set.
• A set M is said to be a positively invariant set if
x(0) ∈ M ⇒ x(t) ∈ M, ∀t ≥ 0
• x(t) approaches a set M as t → ∞, if
∀ > 0 ∃T > 0 : dist(x(t), M ) < , ∀t > T
where dist(p, M ) = inf x∈M kp − xk.
Lemma 4.1
If a solution x(t) of ẋ = f (x) is bounded and belongs to D for t ≥ 0, then its positive
limit set L∗ is a nonempty, compact, invariant set. Moreover, x(t) approaches L∗ as
t → ∞.
17
La Salle’s Theorem
Theorem 4.11 - La Salle’s Theorem
Let
• Ω ⊂ D be a compact positively invariant set.
• V : D → R be a continuously differentiable function such that V̇ (x) ≤ 0 in Ω.
• E = {x ∈ Ω : V̇ (x) = 0}.
• M be the largest invariant set in E.
Then every solution starting in Ω approaches M as t → ∞.
Proof.
(Outline)
- continuity of V (x) - V (x) → a

- Ω is bounded and closed =⇒ - L+ ⊂ Ω
- V̇ (x) = 0 - V (x) = a on L+
From Lemma 4.1 we have that L+ is an invariant set, V̇ (x) = 0 on L+ . Thus, we

conclude that L∗ ⊂ M ⊂ E ⊂ Ω. Since x(t) is bounded, x(t) → L+ as t → ∞ (by
Lemma 4.1). Hence, x(t) → M as t → ∞.
Note that V (x) is not needed to be positive definite.
18
When E is the origin?

Corollary 4.1
Let x = 0 be an equilibrium point of ẋ = f (x). Let V : D → R be a C 1 positive
definite function containing the origin x = 0 such that V̇ ≤ 0 in D. Let
S = {x ∈ D : V̇ = 0} and suppose that no solution can stay identically in S, other
than the trivial solution x(t) = 0. Then, the origin is asymptotically stable.
Corollary 4.2
Let x = 0 be an equilibrium point of ẋ = f (x). Let V : Rn → R be a C 1 , radially
unbounded, positive definite function such that V̇ ≤ 0 for all x ∈ Rn . Let
S = {x ∈ Rn : V̇ = 0} and suppose that no solution can stay identically in S, other
than the trivial solution x(t) = 0. Then x = 0 is GAS.
19
Example
ẋ1 = x2
ẋ2 = −h1 (x1 ) − h2 (x2 )
where
Ry hi (0) = 0 and y hi (y) > 0, ∀y 6= 0, i = 1, 2. Also assume that
0 h1 (z)dz → ∞ as kyk → ∞.
Consider Z x1 1 2
V (x) = h1 (y)dy + x
0 2 2
Clearly V (0) = 0, V (x) > 0, ∀x 6= 0 and V (x) → ∞ as |x| → ∞.
V̇ (x) = h1 (x1 )ẋ1 + x2 ẋ2

= h1 (x1 )x2 − x2 h1 (x1 ) − x2 h2 (x2 )
= −x2 h2 (x2 ) ≤ 0
i.e, negative semidefinite.

Is it GAS?
20
S = {x ∈ R2 : V̇ (x) = 0} = {x ∈ R2 : x2 = 0}
Let x(t) be a solution that belongs identically to S:
x2 (t) = 0 ⇒ ẋ2 = 0 ⇒ x1 = 0
Therefore, the only solution that can stay identically in S is the trivial solution
x(t) = 0. Thus, x = 0 is GAS.
Remark: LaSalle’s theorem extends Lyapunov theorem in three ways besides the one
that we saw above:
1. Can be used in cases where the system has an equilibrium set,
2. It can give an estimate of the region of attraction which is not necessarily of the
form Ωc = {x : V (x) ≤ c},
3. V (x) does not have to be positive definite.
21
Example
ẏ = ay + u
where y ∈ R, a is an unknown but constant parameter and u is the input signal.

Consider the following adaptive control law
u = −ky, k̇ = γy 2 , γ>0
Let x1 = y and x2 = k then the closed loop system is given by
ẋ1 = −(x2 − a)x1

ẋ2 = γx21
We have an equilibrium set Se = {x ∈ R2 : x1 = 0}.
Goal: Show that y → 0, that is, x approaches Se as t → ∞.
22
Goal: Show that y → 0, that is, x approaches Se as t → ∞.
Consider the Lyapunov function candidate

1 2 1
V (x) = x + (x2 − b)2
2 1 2γ
where b > a but we do not know its value explicitly.

1
V̇ (x) = x1 ẋ1 +(x2 − b)ẋ2
γ
1
= −(x2 − a)x21 + (x2 − b)γx21
γ
= −(b − a)x21 ≤ 0.
Let Ωx = {x ∈ R2 : V (x) ≤ c} where c can be chosen large enough since V is radially

unbounded. Let E = {x ∈ R2 : x1 = 0}. By LaSalle’s theorem with r = rc we
conclude that x(t) approaches M = E since E is an invariant set.
23
Linear systems and Linearization
Consider the time-invariant linear system
ẋ = Ax
Theorem 4.6
A matrix A is Hurwitz, that is, Re(λi ) < 0 for all eigenvalues of A, if and only if, for
any given Q = QT > 0 there exists a P = P T > 0 that satisfies the Lyapunov
equation
P A + AT P = −Q (2)
Moreover, if A is Hurwitz, then P is unique solution of (2).
How can we use this result to study the stability of ẋ = f (x) ?
24
Linear systems and Linearization

Goal: Study the stability of ẋ = f (x) using linearization
ẋ = f (x) ⇔ ẋ = Ax + g(x)
˛
∂f (x) ˛
where A = ∂x ˛x=0
and g(x) = f (x) − Ax.
In particular,
∂fi (0)
gi (x) = fi (x) −
∂x
By the mean value theorem
∂fi (zi )
fi (x) = fi (0) + x,
∂x
where zi is a point on the line segment connecting x to the origin.
Therefore » –
∂fi (zi ) ∂fi (0)
gi (x) = (zi ) − (zi ) x
∂x ∂x
and satisfies ‚ ‚
‚ ∂fi (zi ) ∂fi (0) ‚
|gi (x)| ≤ ‚
‚ ∂x (z i ) − (z i )‚ kxk
∂x ‚
∂f
By continuity of ∂x
we can conclude that
kg(x)k
→ 0 as kxk → 0
kxk
25
Lyapunov’s indirect method
Theorem 4.7 - Lyapunov’s indirect method

Let x = 0 be an equilibrium point for the nonlinear system
ẋ = f (x)
where f : D → Rn is C1 and D is a neighborhood of the origin. Let

˛
∂f (x) ˛˛
A=
∂x ˛x=0
Then,
1. The origin is asymptotically stable if Re(λi ) < 0 for all eigenvalues of A.
2. The origin is unstable if Re(λi ) > 0 for one or more of the eigenvalues of A.
Remark that if λi = ˛0, ∀i we cannot say nothing about it. For example
ẋ = ax3 ⇒ A = ∂f
˛
= 3ax3 ˛x=0 = 0. For a = 0, ẋ = 0, which implies that the
˛
∂x ˛ x=0
origin is stable. Taking V = 12 x2 we have V̇ = ax4 and if a < 0 the origin is
asymptotically stable, if a > 0 then it is unstable.
26
Proof
Starting with the proof of (1). Consider the following Lyapunov function candidate
V (x) = xT P x
The derivative of V (x) along the trajectories is given by
V̇ (x) = xT P f (x) + f (x)T P x

= xT P [Ax + g(x)] + [xT AT + g T (x)]P x
= xT (P A + AT P )x + 2xT P g(x)
= −xT Qx + 2xP g(x)
Since A is Hurwitz, xT QX > 0. Regarding the other term, note that
kg(x)k2
→0
kxk2
as kxk2 → 0.
27
Proof
Therefore ∀γ > 0 ∃r > 0 such that
kg(x)k2
< γ, ∀ kxk2 < r
kxk2
which implies that

kg(x)k2 < γkxk2 , ∀ kxk2 < r
Hence,
V̇ (x) < −xT Qx + 2γkP k2 kxk22 , ∀kxk2 < r
Using the fact that
0 < λmin (Q)kxk22 ≤ xT Qx ≤ λmax (Q)kxk22
it follows that
V̇ (x) < −[λmin (Q) − 2γkP k2 ]kxk22 , ∀kxk2 < r
Choosing γ such that λmin (Q) > 2γkP k2 , that is,
λmin (Q)
γ< ⇒ V̇ (x) < 0
2kP k2
and therefore x = 0 is asymptotically stable.
28
Comparison Functions
Definition
A continuous function α : [0, a) → [0, ∞) is said to belong to class K if it is strictly
increasing and α(0) = 0. It is of class K∞ if a = ∞ and α(r) → ∞ as r → ∞.
Definition
A continuous function β : [0, a) × [0, ∞) → [0, ∞) is said to belong to class KL if, for
each fixed t, the mapping β(r, t) belong to class K with respect to r and, for each
fixed r, the mapping β(r, t) is decreasing with respect to t and β(r, t) → 0 as t → ∞.
29
Examples
• Class K∞
α(r) = rc , c>0
0 c−1
α (r) = cr > 0, lim α(r) = ∞
r→∞
• Class K
α(r) = tan−1 (r)
• Class KL
1.
c −t
β(r, t) = r e , c>0
2.
r
β(r, t) = , k > 0,
ktr + 1
∂β 1 ∂β −kr 2
= > 0, = <0
∂r (ktr + 1)2 ∂t (ktr + 1)2
so β(r, t) → 0 as t → ∞.
30
Properties
Lemma (Properties)
Let α1 , α2 ∈ K on [0, a), α3 , α4 ∈ K∞ and β ∈ KL and α−1
i denotes the inverse of
αi :
1. α−1
1 is defined on [0, α1 (a)) and belongs to class K
2. α−1
3 is defined on [0, ∞) and belongs to class K∞
3. α1 ◦ α2 belongs to class K
4. α3 ◦ α4 belongs to class K∞
5. σ(r, t) = α1 (β(α2 (r), t)) belongs to class KL
31
Lemma 4.3
Let V : D → R be a continuous positive definite function defined on a domain
D ⊂ Rn that contains the origin. Let Br ⊂ D for some r > 0. Then, there exist
α1 , α2 ∈ K defined on [0, a] such that
α1 (kxk) ≤ V (x) ≤ α2 (kxk), ∀x ∈ Br
If D = Rn , α1 and α2 are defined on [0, ∞). Moreover, if V (x) is radially unbounded,

then α1 , α2 ∈ K∞ .
Example: Consider a quadratic positive definite function
V (x) = xT P x
We know that
λmin (P )kxk22 ≤ xT P x ≤ λmax (P )kxk22
and so we can define α1 (r) = λmin (P )r2 and α2 (r) = λmax (P )r2
32
Lemma 4.4
Consider the scalar autonomous differential equation
ẏ = −α(y), y(t0 ) = y0
where α is a locally Lipschitz class K function defined on [0, a). Then, for all
0 ≤ y0 < a, the solution is unique and defined for all t ≥ t0 . Moreover,
y(t) = β(y0 , t − t0 )
where β ∈ KL is defined on [0, a) × [0, ∞).

a) ẏ = −ky, k > 0 the solution is given by y(t) = y0 e−k(t−t0 ) and we can define
β(r, t) = re−kt
y0
b) ẏ = −ky 2 , k > 0 the solution is given by y(t) = ky0 (t−t0 )+1
and we can define
r
β(r, t) = krt+1
33
Alternative proof of Lyapunov Theorem (Theorem 4.1)
In the proof, given an > 0, we choose r ≤ such that Br ⊂ D. Now, we would like
to choose β and δ such that
Bδ ⊂ Ωβ ⊂ Br
But V (x) satisfies
α1 (kxk) ≤ V (x) ≤ α2 (kxk)
Therefore select
β ≤ α1 (r)
Why?
V (x) ≤ β ≤ α1 (r) ⇒ α1 (kxk) ≤ α1 (r) ⇔ kxk ≤ r
so we conclude that ∀x ∈ Ωβ ⇒ x ∈ Br .
Also select
δ ≤ α−1
2 (β)
Why?
kxk ≤ δ ⇒ V (x) ≤ α2 (δ) ≤ β ⇒ Bδ ⊂ Ωβ
34
Alternative proof of Lyapunov Theorem
To prove asymptotic stability note that
V̇ (x) < 0 ⇒ ∃α3 , α4 ∈ K : α3 (kxk) ≤ −V̇ (x) ≤ α4 (kxk)
In particular we have that V̇ (x) ≤ −α3 (kxk). Hence
V̇ ≤ −α3 (α−1
2 (V ))
Using the comparison lemma it follows that V ≤ y for
ẏ = −α3 (α−1
2 (y)), y(0) = V (x(0))
From Lemma 4.2 it follows that α3 ◦ α−1

2 ∈ K and from Lemma 4.4 we conclude that
y(t) = β(y(0), t), β ∈ KL
In conclusion V (x(t)) ≤ β(V (x(0)), t), which shows that V is bounded and V → 0 as
t → ∞.
35
Alternative proof of Lyapunov Theorem
In fact we can estimate a bound for x(t)

From
V (x(t)) ≤ V (x(0)) ⇒ α1 (kx(t)k) ≤ V (x(t)) ≤ V (x(0)) ≤ α2 (kx(0)k)
and therefore,
kx(t)k ≤ α−1
1 (α2 (kx(0)k))
where α−1
1 ◦ α2 ∈ K
Similarly, from
V (x(t)) ≤ β(V (x(0)), t) ⇒ α1 (kxk) ≤ V (x(t)) ≤ β(V (x(0)), t) ≤ β(α2 (kx(0)k), t)
we obtain
kxk ≤ α−1
1 (β(α2 (kx(0)k), t)) := σ(kx(0)k, t)
with σ ∈ KL.
36
Nonautonomous Systems
Suppose we would like to analyze the stability behavior of the solution ȳ(t) of the
system
ẏ = g(t, y)
How can we do this?
Define
x(t) = y(t) − ȳ(t)
Then, we have
ẋ = g(t, y) − ȳ˙ = g(t, x(t) + ȳ(t)) − ȳ(t)

˙ := f (t, x)
and therefore we obtain the nonautonomous system
ẋ = f (t, x)
with f (t, 0) = 0, ∀t ≥ 0.
This means that we only need to study the stability of the equilibrium point x = 0.
37
Consider the nonautonomous system
ẋ = f (t, x)
Definition
The origin is an equilibrium point at t = 0 if
f (t, 0) = 0, ∀t ≥ 0.
38
Consider the nonautonomous system
ẋ = f (t, x)
Definition
The equilibrium point x = 0 of the nonautonomous system is
• Stable if ∀ > 0 ∃δ = δ(, t0 ) > 0 : kx(t0 )k < δ ⇒ kx(t)k < , ∀t ≥ t0 ≥ 0
• Uniformly Stable (US) if δ is independent of t0
• Asymptotically Stable if it is stable and ∃c = c(t0 ) : x(t) → 0 as t → ∞

∀kx(t0 )k < c
• Uniformly Asymptotically Stable (UAS) if c is independent of t0 and the

convergence is uniformly in t0 , that is
∀η > 0 ∃T = T (η) > 0 : kx(t)k < η, ∀t ≥ t0 + τ (η), ∀kx(t0 )k < c
• Globally Uniformly Asymptotic Stable (GUAS) if δ() can be chosen to satisfies

lim→∞ δ() = ∞ and
∀(η, c) ∃T = T (η, c) > 0 : kx(t)k < η, ∀t ≥ t0 + T (η, c), ∀kx(t0 )k < c
39
The following lemma make the above definitions more clear.

Lemma 4.5
The equilibrium point x = 0 of the nonautonomous system is
• US ⇔ there exist α ∈ K and c > 0 (independent of t0 ) such that
kx(t)k ≤ α(kx(t0 )k), ∀t ≥ t0 ≥ 0, ∀kx(t0 )k < c
• UAS ⇔ there exist β ∈ KL and c > 0 (independent of t0 ) such that
kx(t)k ≤ β(kx(t0 )k, t − t0 ), ∀t ≥ t0 ≥ 0, ∀kx(t0 )k < c
• GUAS ⇔ the inequality in UAS holds for all x(t0 ), that is c = ∞
40
Definition
The equilibrium point x = 0 is exponentially stable if there exists positive constants
c, k and λ such that
kx(t)k ≤ kkx(t0 )ke−λ(t−t0 ) , ∀kx(t0 k < c
and globally exponentially stable if this inequality is satisfied ∀x(t0 )
Theorem 4.8
Let x = 0 be an equilibrium point that belongs to D ⊂ Rn . Let V : [0, ∞) × D → R
be a C 1 function such that
W1 (x) ≤ V (t, x) ≤ W2 (x)
and
∂V ∂V
+ f (t, x) ≤ 0, ∀t ≥ 0 ∀x ∈ D
∂t ∂x
where W1 (x), W2 (x) are continuous positive functions on D. Then, x = 0 is
uniformly stable.
41
Proof
Choose r > 0 and c > 0 such that Br ⊂ D and

0 < c < minkxk=r W1 (x) ⇒ {x ∈ Br : W1 (x) ≤ c} ⊂ Br
Define a time-dependent set Ωt,c such that
{x ∈ Br : W2 (x) ≤ c} ⊂ Ωt,c = {x ∈ Br : V (t, x) ≤ c} ⊂ {x ∈ Br : W1 (x) ≤ c}
Hence the solution is bounded and defined for all t ≥ t0 . Moreover, since V̇ ≤ 0 we
have that V (t, x(t)) ≤ V (t0 , x(t0 )), ∀t ≥ t0 .
Also
∃α1 ,α2 ∈K : α1 (kxk) ≤ W1 (x) ≤ V (t, x) ≤ W2 (x) ≤ α2 (kxk)
and therefore,
kx(t)k ≤ α−1 −1 −1
1 (V (t, x(t))) ≤ α1 (V (t0 , x(t0 ))) ≤ α1 (α2 (kx(t0 )k))
Thus, we can conclude that x = 0 is US since α−1

1 ◦ α2 ∈ K.
42
Theorem 4.9
Same assumptions as Theorem 4.8 but with
∂V ∂V
+ f (t, x) ≤ −W3 (x), ∀t ≥ 0 ∀x ∈ D
∂t ∂x
where W3 (x) is a continuous positive definite function on D. Then, x = 0 is UAS.
Moreover, if r and c are such that
Br = {kxk ≤ r} ⊂ D
and
c < min W1 (x)
kxk≤r
Then
∀x0 ∈ {x ∈ Br : W2 (x) ≤ c}, kx(t)k ≤ β(kx(t0 )k, t − t0 ), ∀t ≥ t0 ≥ 0
for some β ∈ KL.

If in addition D = Rn and W1 (x) is radially unbounded then x = 0 is GUAS.
43
Proof
∂V ∂V
V̇ (t, x) = + f (t, x) ≤ −W3 (x) ≤ −α3 (kxk),
∂t ∂x
for some α3 ∈ K,
V ≤ α2 (kxk) ⇔ α−1 −1
2 (V ) ≤ kxk ⇔ α3 (α2 (V )) ≤ α3 (kxk)
Thus,
V̇ ≤ −α3 (α−1
2 (V )) := −α(V ), (3)
where α ∈ K.
Without loss of generality α is locally Lipschitz. Why? If not we can always choose
γ ∈ K, with γ Lipschitz such that α(r) ≥ γ(r). Then V̇ ≤ −γ(V ).
44
Proof
√
For example, suppose α(r) = r is a class K function, but not locally Lipschitz at
r = 0. Define 
r,
√ r<1
γ (r) =
r, r ≥ 1
and so, γ ∈ K locally Lipschitz and α(r) ≥ γ(r), ∀r ≥ 0.
Returning to (3) and resorting to the comparison lemma,
ẏ = −α(y), y(t0 ) = V (t0 , x(t0 )) ≥ 0 ⇒ V (t, x(t)) ≤ y(t), ∀t ≥ t0
By Lemma 4.4, ∃σ ∈ KL such that
V (t, x(t)) ≤ σ(V (t0 , x(t0 )), t − t0 )
In conclusion, ∀x(t0 ) ∈ {x ∈ Br : W2 (x) ≤ c} we have
kxk ≤ α−1
1 (V (t, x(t))) ≤ σ(V (t0 , x(t0 )), t−t0 ) ≤ σ(α2 (kx(t0 )k), t−t0 ) =: β(kx(t0 )k, t−t0 )
45
Definition
• V (t, x) is said to be positive semidefinite if V (t, x) ≥ 0
• V (t, x) is said to be positive definite if V (t, x) ≥ W1 (x), for some W1 (x) positive
definite function.
• V (t, x) is radially unbounded if W1 (x) is radially unbounded.
Theorem 4.10
Let x = 0 be an equilibrium point for the nonautonomous system ẋ = f (t, x) that
belongs to some D ⊂ Rn . Let V : [0, ∞) × D → R be a C 1 function such that
k1 kxka ≤ V (t, x) ≤ k2 kxka
∂V ∂V
+ f (t, x) ≤ −k3 kxka , ∀t ≥ 0, ∀x ∈ D
∂t ∂x
where k1 , k2 , k3 and a are positive constants. Then x = 0 is exponentially stable. If
the assumption holds globally. Then x = 0 is GES.
46
proof
k
k3 − 3 (t−t0 )
V̇ ≤ − V ⇒ V (t, x(t)) ≤ V (t0 , x(t0 ))e k2
k2
Hence,
2 31/a
k
− 3 (t−t0 )
V (t,x) 1/a k2
h i
V (t0 ,x(t0 ))e
kxk ≤ k1
≤4 k1
5
2 31/a
k3
a − k (t−t0 ) h ia k
− k 3a (t−t0 )
≤ 4 k2 kx(t0 )k e 2 5 ≤ k2
kx(t0 )ke 2
k1 k1
47
Examples
Example 1
ẋ = −(1 + g(t))x3
where x ∈ R2 , g is C 0 (continuous) and g(t) ≥ 0, ∀t ≥ 0.
Consider
1 2
V = x
2
and so,
V̇ = −(1 + g(t))x4 ≤ −x4
Then, W1 (x) = W2 (x) = V (x) and W3 (x) = x4 , thus x = 0 is GUAS.
Note we cannot conclude exponential because is not the same a.
48
Example 2
ẋ1 = −x1 − g(t)x2
ẋ2 = x1 − x2
where g is C 1 , 0 ≤ g(t) ≤ k and ġ(t) ≤ g(t)
Consider
V (t, x) = x21 + (1 + g(t))x22
Note that,
x21 + x22 ≤ V (t, x) ≤ x21 + (1 + k)x22
Thus V (t, x) is positive definite and radially unbounded.
V̇ (t, x) = 2x1 ẋ1 + 2x2 (1 + g(t))ẋ2 + ġx22

= −2x21 − 2gx1 x2 + 2x2 x1 − 2x22 + 2x2 gx1 − 2x22 g + ġx22
= −2x21 + 2x2 x1 − [2 + 2g − ġ]x22
≤ −2x21 + 2x1 x2 − 2x22
» –» –
2 −1 x1
=: −xT Qx
ˆ ˜
= − x1 x2
−1 2 x2
Therefore, W1 , W2 and W3 are positive definite quadratic functions (a=2)

(λmin (P )xT x ≤ xT P x ≤ λmax (P )xT x) and so we conclude that x = 0 is GES.
49
Linear time-varying systems and linearization
ẋ(t) = A(t)x(t)
the solution can be represented as
x(t) = Φ(t, t0 )x(t0 )
From linear system theory...

Theorem 4.11
The equilibrium point x = 0 of the linear system ẋ(t) = A(t)x(t) is GUAS if and only
if
kΦ(t, t0 )k ≤ ke−λ(t−t0 ) , ∀t ≥ t0 ≥ 0
for some positive constants k and λ.
50
Remark that
1. for linear systems GUAS ⇔ Exponential stability
2. for linear time-varying system, GUAS cannot be characterized by the location of
the eigenvalues of A.
Example
−1 + 1.5cos2 t
» –
1 − 1.5 sin t cos t
A(t) =
1 − 1.5 sin t cos t −1 + 1.5sin2 t
√
For each t, λi (A(t)) = −0.25 ± 0.25 7j. Yet, the origin is unstable. Φ(t, 0) is given
by
e−t sin t
» 0.5t –
e cos t
Φ (t, 0) = 0.5t −t
−e sin t e cos t
51
Theorem 4.12
Let x = 0 be a exponential stable equilibrium point of the linear system
ẋ(t) = A(t)x(t). Suppose A(t) is continuous and bounded. Let Q(t) = QT (t) > 0 be
continuous and bounded. Then, there is a P (t) = P T (t) ∈ C 2 bounded matrix that
satisfies
−Ṗ (t) = P (t)A(t) + AT (t)P (t) + Q(t).
Furthermore, V (t, x) = xT P (t)x is a Lyapunov function that satisfies
c1 kxk22 ≤ xT P (t)x ≤ c2 kxk22
and
∂V ∂V
+ A(t)x ≤ −c3 kxk22
∂t ∂x
52
We are now ready for the following theorem:
Theorem 4.13/4.15
ẋ = f (t, x) (4)
where f : [0, ∞) × D → Rn is C 1 , D = {x ∈ R2 : kxk2 < r}, and the Jacobian

matrix ∂f
∂x
is bounded and Lipschitz on D, uniformly in t. Let
˛
∂f ˛
A(t) = (t, x)˛˛
∂x x=0
Then, the origin is an exponentially stable equilibrium point of (4) if and only if it is
an exponentially stable equilibrium point for the linear system ẋ = A(t)x.
53
Proof
Since the Jacobian is bounded
∂fi ∂fi
k (t, x1 ) − (t, x2 )k2 ≤ L1 kx1 − x2 k2 , ∀x1 , x2 ∈ D
∂x ∂x
By the mean value theorem
∂fi
fi (t, x) = fi (t, 0) + (t, zi )x
∂x
where fi (t, 0) = 0 and zi is a point of the line segment connecting x to the origin.
Thus, » –
∂fi ∂fi ∂fi
fi (t, x) = (t, 0)x + (t, zi ) − (t, 0) x
∂x ∂x ∂x
Hence,
f (x, t) = A(t)x + g(t, x)
h i
where A(t) = ∂f
∂x
i
(t, 0) and g(t, x) = ∂f ∂x
i
(t, zi ) − ∂f
∂x
i
(t, 0) x. Thus, we have
n ‚
!1/2
‚ ∂fi ∂fi
X ‚2
kg(t, x)k2 ≤ (t, zi ) − (t, 0)‚ kxk2 ≤ Lkxk2
‚
‚
i=1
∂x ∂x 2
√
where L = nL1 .
54
Proof
V (t, x) = xT P (t)x
V̇ (t, x) = xT P (t)f (t, x) + f T (t, x)P (t)x + xT Ṗ (t)x

= xT [P (t)A(t) + AT (t)P (t) + Ṗ (t)]x + 2xT P (t)g(t, x)
= −xT Q(t)x + 2xT P (t)g(t, x)
≤ −c3 kxk22 + 2c2 Lkxk32
≤ −(c3 − 2c2 Lρ)kxk22 , ∀kxk2 < ρ
c3
Choosing ρ < min{r, 2c2 L
}, implies that V̇ (t, x) < 0, ∀kxk2 < ρ and thus x = 0 is
exponentially stable.
55
Converse Theorems
Theorem 4.16
ẋ = f (t, x)
where f : [0, ∞) × D → Rn
is C1,
D = {x ∈ R2 : kxk < r} and the Jacobian matrix
∂f
∂x
is bounded on D, uniformly in t. Let β ∈ KL and r0 > 0 and that β(r0 , 0) < r.
Let D0 = {x ∈ Rn : kxk < r0 }. Suppose that
kx(t)k ≤ β(kx(t0 )k, t − t0 ), ∀x(t0 ) ∈ D0 ∀t ≥ t0 ≥ 0.
Then, there is a C 1 function V : [0, ∞) × D0 → R that satisfies
α1 (kxk) ≤ V (t, x) ≤ α2 (kxk)
∂V ∂V
+ f (t, x) ≤ −α3 (kxk)
∂t ∂x
‚ ∂V ‚
‚ ≤ α4 (kxk),
‚ ‚
‚
∂x
where α1 , α2 , α3 , α4 are class K functions defined on [0, r0 ]. If the system is
autonomous, V can be chosen independently of t.
In particular, for β(r, t) = kre−λt we have αi = ci r2 , i = 1, 2, 3, α4 = c4 r together
with Z t+δ
V (t, x) = ΦT (τ ; t, x)Φ(τ ; t, x)dτ
t
where Φ(τ ; t, x) denotes the solution of the system that starts at (t, x).
56
Boundedness and Ultimate Boundedness
Until now we have used Lyapunov theory to study the behavior of the system about
the equilibrium point.
What happens when the system does not have any equilibrium point?
We will see that Lyapunov analysis can be used to show boundedness of the solution
of the state equation.
Example
ẋ = −x + δ sin t, x(t0 ) = a, a > δ > 0
There are no equilibrium points!
Nevertheless, with
1
V (x) =
2
we obtain
V̇ = −x2 + xδ sin t ≤ −x2 + δ|x|.
δ2
Note that V̇ < 0, ∀|x| > δ, which means that the set {x ∈ R : V (x) ≤ c} with c > 2
is an invariant set because V = c → V̇ < 0.
Hence, the solutions are uniformly bounded.
57
Boundedness and Ultimate Boundedness

Moreover
δ2
< < c −→ V̇ < −γ
∀ > 0 :
2
for some γ > 0 in the set ≤ V ≤ c which shows that V will reach in finite time
and the solution enter the set {V ≤ }.
Thus, we can√conclude that the solution is uniformly ultimately bounded with ultimate
bound |x| ≤ 2.
Definition
The solutions of ẋ = f (t, x) are
• Uniformly Bounded (UB) if there exists a c > 0, independent of t0 ≥ 0 such that
∀a ∈ (0, c) ∃β = β(a) > 0 : kx(t0 )k ≤ a ⇒ kx(t)k ≤ β, ∀t ≥ t0
• Globally Uniformly Bounded (GUB) if a can be arbitrarily large

• Uniformly Ultimately Bounded (UUB) with ultimate bound b, if there exists
b > 0, c > 0 (independent of t0 ) such that
∀a ∈ (0, c) ∃T = T (a, b) ≥ 0 : kx(t0 )k ≤ a ⇒ kx(t)k ≤ b, ∀t ≥ t0 + T,
• Globally Uniformly Ultimately Bounded (GUUB) if a can be arbitrarily large.
58
Theorem 4.18 (Global)

Let V : [0, ∞) × Rn → R be a C 1 function such that
α1 (kxk) ≤ V (t, x) ≤ α2 (kxk)
∂V ∂V
+ f (t, x) ≤ −W3 (x), ∀kxk ≥ µ > 0
∂t ∂x
∀t ≥ 0 and ∀x ∈ Rn , where α1 , α2 ∈ K∞ and W3 (x) > 0.
Then, there exists a β ∈ KL and T ≥ 0 such that
kx(t)k ≤ β(kx(t0 )k, t − t0 ), ∀t0 ≤ t ≤ t0 + T
and
kx(t)k ≤ α−1
1 (α2 (µ)), ∀t ≥ t0 + T
59
Example - Mass-spring system

mÿ + cẏ + ky + ka2 y 3 = F = A cos(wt)
Taking
x1 = y
x2 = ẏ
and assuming certain numerical values we get
ẋ1 = x2
ẋ2 = −(1 + x21 )x1 − x2 + M cos(wt)
where M ≥ 0 is proportional to A. Choosing
» 3
/2 1/2
–
V = xT 1/ x + 1/2x41
2 1
then
V̇ = −x21 − x41 − x22 + (x1 + 2x2 ) M cos (wt)
» –
˜ x1
= −kxk2 − x41 + 1 2
ˆ
M cos (wt)
x2
√
≤ − kxk2 − x41 + M 5 kxk
√
≤ − (1 − θ) kxk2 − x41 − θ kxk2 + M 5 kxk , 0 < θ < 1
60
Example - Mass-spring system

Therefore √
2 M 5
V̇ ≤ − (1 − θ) kxk − x41 ,
∀kxk ≥
θ
which shows that the solutions are GUUB. To compute the ultimate bound, we have
to find α1 , α2 .
V (x) ≥ xT P x ≥ λmin (P )kxk2 → α1 (r) = λmin (P )r2
1 1 1 4
V (x) ≤ xT P x + kxk4 ≤ λmax (P )kxk2 + kxk4 → α2 (r) = λmax (P )r2 + r
2 2 2
√
M 5
Let µ = θ
kxk ≤ µ ⇒ V (x) ≤ α2 (µ) =
which ensures that
Bµ ⊂ Ω = {x : V (x) ≤ }
But
V (x) ≤ ⇒ α1 (kxk) ≤ ⇔ kxk ≤ α−1
1 ()
therefore v
µ4
u
u λmax (P )µ2 +
x ∈ Ω ⇒ kxk ≤ α−1
1 (α2 (µ)) = 2
t
λmin (P )
61
Input-to-state stability
For a linear time-invariant system
ẋ = Ax + Bu
with A Hurwitz. The solution is given by

Z t
x(t) = eA(t−t0 ) x(t0 ) + eA(t−τ ) Bu(τ )dτ
t0
Using the bound

keA(t−t0 ) k ≤ ke−λ(t−t0 )
we conclude that
Z t
kx(t)k ≤ ke−λ(t−t0 ) kx(t0 )k + ke−λ(t−τ ) kBkku(τ )kdτ
t0
k
≤ ke−λ(t−t0 ) kx(t0 )k + (1 − e−λ(t−t0 ) )kBk sup ku(τ )k
λ t0 ≤τ ≤t
k
≤ ke−λ(t−t0 ) kx(t0 )k + kBk sup ku(τ )k
λ t0 ≤τ ≤t
Does this hold for a general nonlinear system?
ẋ = f (t, x, u)
And in what conditions? Is it sufficient to have GUAS of the unforced system? NO!
62
Input-to-state stability
Example
ẋ = −3x + (1 + 2x2 )u
when u = 0, the equilibrium point x = 0 is GAS. Yet, when x(0) = 2 and u(t) = 1,
the solution is given by
3 − et
x(t) =
3 − 2et
which is unbounded. It even has a finite escape time.
Definition
The system
ẋ = f (t, x, u)
is said to be input-to-state stable (ISS) if there exist β ∈ KL, and γ ∈ K such that for
any initial state x(t0 ) and any bounded input u(t), the solution x(t) exists for all
t ≥ t0 and satisfies
kx(t)k ≤ β(kx(t0 )k, t − t0 ) + γ( sup ku(τ )k)

t0 ≤τ ≤t
63
Remarks:
1. For any bounded input u(t), the state x(t) will be bounded
2. As t increases, the state x(t) will be ultimately bounded by a class K function of
supt≥t0 ku(t)k
3. If u(t) → 0 as t → ∞ then x(t) → 0 as t → ∞
4. Since for u(t) = 0, kx(t)k ≤ β(kx(t0 )k, t − t0 ) ⇔ x = 0 of the unforced system
is GUAS.
Theorem 4.19
Let V : [0, ∞) × Rn → R be a C 1 function such that
α1 (kxk) ≤ V (t, x) ≤ α2 (kxk)
∂V ∂V
+ f (t, x, u) ≤ −W3 (x), ∀kxk ≥ ρ(u) > 0
∂t ∂x
∀(t, x, u) ∈ [0, ∞) × Rn × Rn , α1 , α2 ∈ K∞ , ρ ∈ K, W3 (x) > 0 and C 0 . Then the
system ẋ = f (t, x, u) is ISS with γ = α−1
1 ◦ α2 ◦ ρ.
Proof.
Apply Theorem 4.18.
For autonomous systems the conditions are also necessary, that is, it is if and only if.
64
Lemma 4.6
Suppose that f (t, x, u) is C 1 and globally Lipschitz in (x, u), uniformly in t. If the
unforced system has a globally exponentially stable equilibrium point at x = 0, then
the system ẋ = f (t, x, u) is ISS
Proof.
From the converse theorem (Theorem 4.14) we conclude that the unforced system has
a Lyapunov function V (t, x). Thus,
∂V ∂V ∂V
V̇ = + f (t, x, 0) + [f (t, x, u) − f (t, x, 0)]
∂t ∂x ∂x
and therefore
V̇ ≤ −c3 kxk2 + c4 kxkLkuk ≤ −c3 (1 − θ)kxk2 − c3 θkxk2 + c4 Lkxkkuk
where 0 < θ < 1. Then,
c4 Lkuk
V̇ ≤ −c3 (1 − θ)kxk2 , ∀kxk ≥
c3 θ
Apply Theorem 4.19 with
c4 L
α1 (r) = c1 r2 , α2 (r) = c2 r2 , ρ(r) = r
c3 θ
q
c2 c4 L
we conclude that the system is ISS with γ(r) = c1 c3 θ
r
65
Example
ẋ = −x3 + u
Choose,
1 2
V = x
2
Then
V̇ = −x4 + xu
= −(1 − θ)x4 − θx4 + xu, 0 < θ < 1
|u| 1/3
„ «
≤ −(1 − θ)x4 , ∀|x| ≥
θ
` r ´1/3
which means that the system is ISS with γ = θ
and γ = α−1
1 ◦ α2 ◦ ρ.
66
Cascade System
Consider the following system
ẋ1 = f1 (t, x1 , x2 ) (5)
where the equilibrium point x1 = 0 of ẋ1 = f1 (t, x1 , 0), is GUAS in cascade with
ẋ2 = f1 (t, x2 ), (6)
where the equilibrium point x2 = 0 of ẋ2 = f1 (t, x2 ), is GUAS.

Under what conditions will the origin x = (x1 , x2 ) = 0 of the cascade system be
GUAS?
Lemma 4.7
If system (5) is ISS with x2 as input, then the origin of the cascade system is GUAS.
67
Proof
kx1 (t)k ≤ β1 (kx1 (s)k, t − s) + γ1 ( sup kx2 (τ )k) (7)

s≤τ ≤t
kx2 (t)k ≤ β2 (kx1 (s)k, t − s) (8)

t+t0
Taking (7) with s = 2
yields
t + t0 t − t0
kx1 (t)k ≤ β1 (kx1 ( )k, ) + γ1 ( sup kx2 (τ )k)
2 2 t+t0
≤τ ≤t
2
but
t + t0 t − t0
kx1 ( )k ≤ β1 (kx1 (t0 )k, ) + γ1 ( sup kx2 (τ )k).
2 2 t+t0
t 0 ≤τ ≤ 2
From (8)
sup kx2 (τ )k ≤ β2 (kx2 (t0 )k, 0)
t+t0
t0 ≤τ ≤ 2
t − t0
sup kx2 (τ )k ≤ β2 (kx2 (t0 )k, )
t+t0
≤τ ≤t
2
2
68
Proof
Thus,
t − t0
kx1 (t)k ≤ β1 (β1 (kx1 (t0 )k, )
2
t − t0
+ γ1 (β2 (kx2 (t0 )k, 0)), )
2
t − t0
+ γ1 (β2 (kx2 (t0 )k, ))
2
Since kx(t)k ≤ kx1 (t)k + kx2 (t)k and kx1 (t0 )k ≤ kx(t0 )k, kx2 (t0 )k ≤ kx(t0 )k then
kx(t)k ≤ β(kx(t0 )k, t − t0 ),
where
β(r, s) = β1 (β1 (r, s/2) + γ1 (β2 (r, 0)), s/2) + γ1 (β2 (r, s/2)) + β2 (r, s)
69
ISS Small-Gain Theorem
Consider the interconnected system
ẋ1 = f1 (t, x1 , x2 , w1 )
ẋ2 = f2 (t, x1 , x2 , w2 )
Assume that f1 (t, 0, 0, 0) = f2 (t, 0, 0, 0, 0) = 0, ∀t ≥ t0 ≥ 0 where

f1 (t, ., ., .), f2 (t, ., ., .) are piecewise continuous in t and locally Lipschitz in the rest of
arguments. Suppose that
• x1 subsystem is ISS with respect to state x1 and inputs (x2 , w1 ), that is
∃β1 ∈KL ∃γ1 ,γw1 ∈K such that for all x1 (t0 ) and any bounded (x2 , w1 )
kx1 (t)k ≤ β1 (kx1 (t0 )k, t − t0 ) + γ1 ( sup kx2 (τ )k) + γw1 ( sup kw1 (τ )k)
τ ∈[t0 ,t] τ ∈[t0 ,t]
• x2 subsystem is ISS with respect to state x2 and inputs (x1 , w2 )

• γ1 ◦ γ2 (r) < r ∀r > 0 (contraction condition)
Then, the interconnect system is ISS with respect to state x = [x1 x2 ]T and input
w = [w1 w2 ]T
70

Dynamic Security - Assessment

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Dynamic Security - Assessment

Diunggah oleh

Hak Cipta:

Format Tersedia

4.

Nonlinear Control Systems

António Pedro Aguiar

IST DEEC PhD Course

Goal: Stability analysis of the equilibrium point x̄ ∈ D.

Without loss of generality, we consider that x̄ = 0.

kx(0)k < δ ⇒ kx(t)k < , ∀t ≥ 0

• unstable if it is not stable

kx(0)k < δ ⇒ lim x(t) = 0

where x1 = θ, x2 = θ̇, and β is the friction. The following equilibrium points

V (x) = ax21 + 2x1 x3 + ax22 + 4x2 x3 + ax23

Lyapunov’s stability theorem

Theorem 4.1 - Lyapunov’s stability theorem

• V (0) = 0, V (x) > 0, ∀x ∈ D\ {0}

Then, x = 0 is stable. Moreover, if

V̇ (x) < 0, ∀x ∈ D\ {0}

then x = 0 is asymptotically stable.

• Let α = minkxk=r V (x). Then α > 0.

V̇ (x(t)) ≤ 0 ⇒ V (x(t)) ≤ V (x(0)) ≤ β, ∀t ≥ 0

kx(t)k ≤ δ ⇒ V (x) < β

∀a > 0 ∃τ > 0 : kx(t)k < a, ∀t > τ

Since V is monotonically decreasing and bounded from below by zero, then

because V̇ (x(τ )) ≤ −γ = maxd≤kxk≤r V̇ (x). Thus, V will eventually become

g(0) = 0, xg(x) > 0, ∀x 6= 0, x ∈ (−a, a)

2. Consider the Pendulum example without friction

Assume the following energy function

Clearly, V (0) = 0 and V (x) > 0, −2π < x1 < 2π, x1 6= 0.

V̇ (x) = a sin(x1 )ẋ1 + x2 ẋ2

3. Pendulum equation, but this time with friction

3. However, we know that is asymptotically stable. Let us try

When the region of attraction is Rn ?

c < inf V (x).

then Ωc is bounded only if c < l. In the example

Laypunov Stability - Globally Asymptotically Stability

Theorem 4.2 - GAS

• V (0) = 0 and V (x) > 0, ∀x 6= 0

Then x = 0 is globally asymptotically stable (GAS).

Theorem 4.3 - Instability Theorem

and suppose that V̇ (x) > 0 in U . Then, x = 0 is unstable.

The Invariant Principle

• P is said to be a positive limit point of x(t) if there is a sequence {tn }, with

• A set M is said to be a positively invariant set if

• x(t) approaches a set M as t → ∞, if

∀ > 0 ∃T > 0 : dist(x(t), M ) < , ∀t > T

where dist(p, M ) = inf x∈M kp − xk.

• V : D → R be a continuously differentiable function such that V̇ (x) ≤ 0 in Ω.

• M be the largest invariant set in E.

Then every solution starting in Ω approaches M as t → ∞.

- continuity of V (x) - V (x) → a

From Lemma 4.1 we have that L+ is an invariant set, V̇ (x) = 0 on L+ . Thus, we

Note that V (x) is not needed to be positive definite.

When E is the origin?

V̇ (x) = h1 (x1 )ẋ1 + x2 ẋ2

i.e, negative semidefinite.

where y ∈ R, a is an unknown but constant parameter and u is the input signal.

Let x1 = y and x2 = k then the closed loop system is given by

ẋ1 = −(x2 − a)x1

We have an equilibrium set Se = {x ∈ R2 : x1 = 0}.

Goal: Show that y → 0, that is, x approaches Se as t → ∞.

Goal: Show that y → 0, that is, x approaches Se as t → ∞.

Consider the Lyapunov function candidate

where b > a but we do not know its value explicitly.

Let Ωx = {x ∈ R2 : V (x) ≤ c} where c can be chosen large enough since V is radially

kx(0)k < δ ⇒ kx(t)k < , ∀t ≥ 0

∀ > 0 ∃T > 0 : dist(x(t), M ) < , ∀t > T

• Stable if ∀ > 0 ∃δ = δ(, t0 ) > 0 : kx(t0 )k < δ ⇒ kx(t)k < , ∀t ≥ t0 ≥ 0

• Globally Uniformly Asymptotic Stable (GUAS) if δ() can be chosen to satisfies